基于pytorch的天气图像识别实验
- 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
- 🍖 原作者：K同学啊
Ⅰ Ⅰ Ⅰ Introduction：
- 本文是基于pytorch完成天气识别分类任务的实验记录。
- 实验目标：
- 学习本地读取加载数据
- 熟练构建并使用CNN网络
- 测试集acc达标93%+
- 学习分割数据集
Ⅱ Ⅱ Ⅱ Experiment：
- 数据准备与任务分析：
- 数据集为本地已准备好
- 通过分割数据为测试与训练，训练模型至在测试集上有优良表现。
- 配置环境：
语言环境：python 3.8
编译器: pycharm
深度学习环境：
torch2.11
cuda12.1
torchvision0.15.2a0 - 构建网络：
- 导入数据：
data_dir = './data/'
data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]
import matplotlib.pyplot as plt
from PIL import Image
image_folder = './data/cloudy/'
image_files = [f for f in os.listdir(image_folder) if f.endswith((".jpg", ".png", ".jpeg"))]
fig, axes = plt.subplots(3, 8, figsize=(16, 6))
for ax, img_file in zip(axes.flat, image_files):
img_path = os.path.join(image_folder, img_file)
img = Image.open(img_path)
ax.imshow(img)
ax.axis('off')
plt.tight_layout()
plt.show()
total_datadir = './data/'
train_transforms = transforms.Compose([
transforms.Resize([224, 224]),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
total_data = datasets.ImageFolder(total_datadir,transform=train_transforms)
train_size = int(0.8 * len(total_data))
test_size = len(total_data) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(total_data, [train_size, test_size])
batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
- 网络结构：
import torch.nn.functional as F
class Network_bn(nn.Module):
def __init__(self):
super(Network_bn, self).__init__()
"""
nn.Conv2d()函数：
第一个参数（in_channels）是输入的channel数量
第二个参数（out_channels）是输出的channel数量
第三个参数（kernel_size）是卷积核大小
第四个参数（stride）是步长，默认为1
第五个参数（padding）是填充大小，默认为0
"""
self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=0)
self.bn1 = nn.BatchNorm2d(12)
self.conv2 = nn.Conv2d(in_channels=12, out_channels=12, kernel_size=5, stride=1, padding=0)
self.bn2 = nn.BatchNorm2d(12)
self.pool1 = nn.MaxPool2d(2,2)
self.conv4 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=5, stride=1, padding=0)
self.bn4 = nn.BatchNorm2d(24)
self.conv5 = nn.Conv2d(in_channels=24, out_channels=24, kernel_size=5, stride=1, padding=0)
self.bn5 = nn.BatchNorm2d(24)
self.pool2 = nn.MaxPool2d(2,2)
self.fc1 = nn.Linear(24*50*50, len(classeNames))
def forward(self, x):
x = F.relu(self.bn1(self.conv1(x)))
x = F.relu(self.bn2(self.conv2(x)))
x = self.pool1(x)
x = F.relu(self.bn4(self.conv4(x)))
x = F.relu(self.bn5(self.conv5(x)))
x = self.pool2(x)
x = x.view(-1, 24*50*50)
x = self.fc1(x)
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))
model = Network_bn().to(device)
model
- torch.nn.MaxPool2d():
- kernel_size：最大的窗口大小
- stride：窗口的步幅，默认值为kernel_size
- padding：填充值，默认为0
- dilation：控制窗口中元素步幅的参数
- 训练模型：
训练函数：
def train(dataloader, model, loss_fn, optimizer):
size = len(dataloader.dataset) # 训练集的大小，一共60000张图片
num_batches = len(dataloader) # 批次数目，1875（60000/32）
train_loss, train_acc = 0, 0 # 初始化训练损失和正确率
for X, y in dataloader: # 获取图片及其标签
X, y = X.to(device), y.to(device)
# 计算预测误差
pred = model(X) # 网络输出
loss = loss_fn(pred, y) # 计算网络输出和真实值之间的差距，targets为真实值，计算二者差值即为损失
# 反向传播
optimizer.zero_grad() # grad属性归零
loss.backward() # 反向传播
optimizer.step() # 每一步自动更新
# 记录acc与loss
train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
train_loss += loss.item()
train_acc /= size
train_loss /= num_batches
return train_acc, train_loss
- 测试模型：
测试函数：
def test(dataloader, model, loss_fn):
size = len(dataloader.dataset) # 测试集的大小，一共10000张图片
num_batches = len(dataloader) # 批次数目，313（10000/32=312.5，向上取整）
test_loss, test_acc = 0, 0
# 当不进行训练时，停止梯度更新，节省计算内存消耗
with torch.no_grad():
for imgs, target in dataloader:
imgs, target = imgs.to(device), target.to(device)
# 计算loss
target_pred = model(imgs)
loss = loss_fn(target_pred, target)
test_loss += loss.item()
test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()
test_acc /= size
test_loss /= num_batches
return test_acc, test_loss
- 实验结果及可视化：
编写主函数：
if __name__ == '__main__':
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
data_dir = './data/'
data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]
image_folder = './data/cloudy/'
image_files = [f for f in os.listdir(image_folder) if f.endswith((".jpg", ".png", ".jpeg"))]
fig, axes = plt.subplots(3, 8, figsize=(16, 6))
for ax, img_file in zip(axes.flat, image_files):
img_path = os.path.join(image_folder, img_file)
img = Image.open(img_path)
ax.imshow(img)
ax.axis('off')
plt.tight_layout()
plt.show()
total_datadir = './data/'
train_transforms = transforms.Compose([
transforms.Resize([224, 224]),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
total_data = datasets.ImageFolder(total_datadir, transform=train_transforms)
train_size = int(0.8 * len(total_data))
test_size = len(total_data) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(total_data, [train_size, test_size])
batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
model = Network_bn().to(device)
loss_fn = nn.CrossEntropyLoss()
learn_rate = 1e-4 # 学习率
opt = torch.optim.SGD(model.parameters(), lr=learn_rate)
epochs = 30
train_loss = []
train_acc = []
test_loss = []
test_acc = []
for epoch in range(epochs):
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)
model.eval()
epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
train_acc.append(epoch_train_acc)
train_loss.append(epoch_train_loss)
test_acc.append(epoch_test_acc)
test_loss.append(epoch_test_loss)
print("Using {} device".format(device))
template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%，Test_loss:{:.3f}')
print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
print('Done')
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100
epochs_range = range(epochs)
plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')
plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()
经过30轮训练，可以看到在训练集上有较好的拟合效果，并且也在测试集达到了93的准确率：
Using cuda device
Epoch: 1, Train_acc:58.0%, Train_loss:1.043, Test_acc:66.7%，Test_loss:0.993
Using cuda device
Epoch: 2, Train_acc:79.3%, Train_loss:0.641, Test_acc:83.6%，Test_loss:0.548
Using cuda device
Epoch: 3, Train_acc:85.6%, Train_loss:0.533, Test_acc:84.9%，Test_loss:0.455
Using cuda device
Epoch: 4, Train_acc:85.3%, Train_loss:0.469, Test_acc:88.4%，Test_loss:0.418
Using cuda device
Epoch: 5, Train_acc:86.4%, Train_loss:0.427, Test_acc:89.8%，Test_loss:0.393
Using cuda device
Epoch: 6, Train_acc:88.0%, Train_loss:0.384, Test_acc:90.7%，Test_loss:0.396
Using cuda device
Epoch: 7, Train_acc:89.6%, Train_loss:0.362, Test_acc:91.1%，Test_loss:0.291
Using cuda device
Epoch: 8, Train_acc:89.8%, Train_loss:0.319, Test_acc:90.2%，Test_loss:0.367
Using cuda device
Epoch: 9, Train_acc:90.9%, Train_loss:0.294, Test_acc:92.4%，Test_loss:0.273
Using cuda device
Epoch:10, Train_acc:90.6%, Train_loss:0.301, Test_acc:91.1%，Test_loss:0.272
Using cuda device
Epoch:11, Train_acc:90.4%, Train_loss:0.301, Test_acc:91.1%，Test_loss:0.263
Using cuda device
Epoch:12, Train_acc:91.4%, Train_loss:0.266, Test_acc:92.0%，Test_loss:0.260
Using cuda device
Epoch:13, Train_acc:92.1%, Train_loss:0.266, Test_acc:90.2%，Test_loss:0.349
Using cuda device
Epoch:14, Train_acc:93.0%, Train_loss:0.259, Test_acc:91.6%，Test_loss:0.273
Using cuda device
Epoch:15, Train_acc:93.7%, Train_loss:0.243, Test_acc:91.1%，Test_loss:0.294
Using cuda device
Epoch:16, Train_acc:93.7%, Train_loss:0.251, Test_acc:90.7%，Test_loss:0.559
Using cuda device
Epoch:17, Train_acc:93.8%, Train_loss:0.215, Test_acc:92.0%，Test_loss:0.219
Using cuda device
Epoch:18, Train_acc:92.4%, Train_loss:0.230, Test_acc:92.4%，Test_loss:0.235
Using cuda device
Epoch:19, Train_acc:94.1%, Train_loss:0.194, Test_acc:92.4%，Test_loss:0.226
Using cuda device
Epoch:20, Train_acc:93.7%, Train_loss:0.198, Test_acc:92.9%，Test_loss:0.209
Using cuda device
Epoch:21, Train_acc:95.8%, Train_loss:0.173, Test_acc:93.3%，Test_loss:0.224
Using cuda device
Epoch:22, Train_acc:94.6%, Train_loss:0.174, Test_acc:92.9%，Test_loss:0.284
Using cuda device
Epoch:23, Train_acc:93.7%, Train_loss:0.205, Test_acc:93.3%，Test_loss:0.203
Using cuda device
Epoch:24, Train_acc:94.3%, Train_loss:0.197, Test_acc:91.1%，Test_loss:0.256
Using cuda device
Epoch:25, Train_acc:96.0%, Train_loss:0.165, Test_acc:92.9%，Test_loss:0.200
Using cuda device
Epoch:26, Train_acc:96.1%, Train_loss:0.162, Test_acc:93.3%，Test_loss:0.205
Using cuda device
Epoch:27, Train_acc:96.2%, Train_loss:0.166, Test_acc:91.1%，Test_loss:0.257
Using cuda device
Epoch:28, Train_acc:96.2%, Train_loss:0.153, Test_acc:92.0%，Test_loss:0.229
Using cuda device
Epoch:29, Train_acc:96.9%, Train_loss:0.140, Test_acc:93.8%，Test_loss:0.197
Using cuda device
Epoch:30, Train_acc:96.1%, Train_loss:0.152, Test_acc:93.3%，Test_loss:0.194
Done
进程已结束,退出代码0
Ⅲ Ⅲ Ⅲ Conclusion：
- 通过本次实验学习到了，从本地加载并分割数据集的方法。同时对CNN的搭建和使用有了进一步掌握。
- 对于过拟合问题有了初步的认识，在实验中尝试了不同参数，对于训练集的过度训练很容易导致过拟合，使得模型在测试集的表现反而下降。解决过拟合的方法有许多，这是未来值得我进一步学习的地方。
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
